Microsoft has launched its own AI models, MEI Voice One for expressive speech synthesis and MEI One Preview for text generation, aiming to reduce reliance on external partners like OpenAI. While still mid-tier in benchmarks, these models are accessible for testing on platforms like LM Arena and Copilot Labs, showcasing Microsoft’s growing presence and ambition in the AI landscape.
Microsoft has recently launched its own proprietary AI models, marking a significant step beyond its existing partnership with OpenAI and ChatGPT. The new models include MEI Voice One, a highly expressive and natural speech generation model, and MEI One Preview, a text-based AI model designed to compete with other leading models like Gemini and GPT-5. While Microsoft’s CEO claims these models are among the best globally, official benchmark results have yet to be publicly released. However, users can test MEI One Preview on the LM Arena platform, where it competes against other AI models in real-time comparisons.
LM Arena offers an interactive environment where users can prompt different AI models and compare their responses side-by-side. This setup allows users to choose their preferred output, which then reveals which model generated each response. MEI One Preview is featured among various competitors, including Llama 2 and Phantom, but currently ranks around 13th on the leaderboard, indicating it still has room for improvement compared to top-tier models like Gemini 2.5 Pro and GPT-5. The platform also allows users to request API access to integrate these models into their own applications, although details about the API’s availability and cost remain unclear.
The MEI Voice One model focuses on delivering natural and expressive speech synthesis, with a variety of voice styles such as empathetic, pirate, vampire, cowboy, and sports play-by-play. Users can experiment with these voices on Microsoft’s Copilot Labs website, where the model generates not only spoken text but also extended scripts based on user prompts. The voice quality has been praised for its realism and expressiveness, rivaling popular services like 11 Labs and ChatGPT’s voice capabilities, although some of the more whimsical voice options may not appeal to all users.
Microsoft’s motivation behind developing these proprietary models appears to be a desire for greater independence in AI technology, reducing reliance on external partners like OpenAI. This move comes amid some tension in Microsoft’s relationship with OpenAI and Apple, signaling a strategic shift toward building and controlling its own AI infrastructure. By releasing MEI Voice One and MEI One Preview, Microsoft is positioning itself as a serious contender in the AI space, aiming to gather extensive training data and user feedback to refine its offerings.
In summary, Microsoft’s introduction of MEI Voice One and MEI One Preview represents a notable expansion of its AI capabilities, offering users new tools for both text and voice applications. While the models are still developing and currently rank mid-tier in competitive benchmarks, the ability to test and interact with them via LM Arena and Copilot Labs provides valuable insight into their potential. As Microsoft continues to evolve these models and possibly opens up API access, it will be interesting to see how they stack up against established AI leaders in the near future.